Skip to content

Benchmark

BABILong qa3 0k

Long context benchmarks measure how well models use information in long inputs. Scores are percentages. Higher is better.

Models and systems measured
14
Scores recorded
15
Sources
1
IQ weight
No weight published

Score distribution

50%
Mean
36%
Median
25.4%
Std dev
32%
10th percentile
97%
90th percentile
17%
Min
17%97%
The band covers the middle 80% of scores. The line marks the median.

Results

The source marks these results as descriptive, not capability scores, so we list them by name instead of ranking them.

Model or systemScore
~ ARMT (137M) fine-tuneqa3, 0k· BABILong v0 data docs97%
~ Mamba (130M) fine-tuneqa3, 0k· BABILong v0 data docs97%
~ RMT (137M) fine-tuneqa3, 0k· BABILong v0 data docs97%
ChatGLM3 6Bchatglm3-6b-128k, qa3, 0k, 128k· BABILong v0 data docs33%
GPT-4-0125-previewqa3, 0k· BABILong v0 data docs56%
Llama 2 7bLLaMA-2-7B-32K, qa3, 0k, 32k· BABILong v0 data docs32%
Llama 3 8B InstructMeta-Llama-3-8B-Instruct, qa3, 0k· BABILong v0 data docs33%
Llama3-ChatQA-1.5-8B + RAGqa3, 0k· BABILong v0 data docs17%
Mistral 7B Instruct v0.2Mistral-7b-Instruct-v0.2, qa3, 0k· BABILong v0 data docs36%
Mixtral 8x22B Instruct v0.1Mixtral-8x22B-Instruct-v0.1, qa3, 0k· BABILong v0 data docs53%
Mixtral 8x7B Instruct v0.1Mixtral-8x7B-Instruct-v0.1, qa3, 0k· BABILong v0 data docs34%
Phi-3 Medium 128K InstructPhi-3-medium-128k-instruct, qa3, 0k· BABILong v0 data docs53%
Phi-3 Mini 128K InstructPhi-3-mini-128k-instruct, qa3, 0k· BABILong v0 data docs32%
Yi-34B01-ai/Yi-34B-200k, qa3, 0k, 200k· BABILong v0 data docs35%
Yi-34B01-ai/Yi-34B, qa3, 0k· BABILong v0 data docs45%

Data credits: RMT-team, Kuratov et al. Official v0 numeric CSV; results licence not separately stated.